How to show that $\dfrac{a^3}{a^2+b^2} + \dfrac{b^3}{b^2+c^2} + \dfrac{c^3}{c^2+a^2} \ge \dfrac32$, where $a^2+b^2+c^2=3$, and $a,b,c > 0$ ?
-
Haven't tried it, but it seems like a typical multivariable calculus optimization problem: use Lagrange multipliers, consider behavior on the boundary, etc. – Nate Eldredge Apr 13 '14 at 17:25
-
@Nate I will be grateful if the solution involves only pre-calculus or single variable calculus .. L-Multipliers are beyond my scope – r9m Apr 13 '14 at 17:31
-
2Don't see an easy precalc solution. You could set $b=a+x, c=a+y$, expand out $LHS-RHS$ (very messy) and note the resulting expression is non-negative. You may also want to check http://www.artofproblemsolving.com/Forum/viewtopic.php?p=3280223#p3280223 which is the homogeneous form of your inequality. – Macavity Apr 14 '14 at 04:49
-
@Macavity Thanks a ton :) !! – r9m Apr 14 '14 at 08:59
3 Answers
As has been mentioned in some comments too, this question does not seem to allow a pre-calculus solution. Anyway, it is again a question of the symmetric type, such as listed in:
Completed with: From the latter article comes the followingTheorem (The Purkiss Principle). Let $f$ and $g$ be symmetric functions with continuous second derivatives in the neighborhood of a point $P = (r, \cdots, r)$. On the set where $g$ equals $g(P)$, the function $f$ will have a local maximum or minimum at $P$ except in degenerate cases.
Our function $f$ in this case is: $$ f(a,b,c) = \frac{a^3}{a^2+b^2}+\frac{b^3}{b^2+c^2}+\frac{c^3}{c^2+a^2}$$ Applying the Purkiss Principle gives (with $r > 0$): $$ g(r,r,r)=r^2+r^2+r^2=3 \quad \Longrightarrow \quad r=1 \quad \Longrightarrow \quad f(r,r,r) = \frac{3}{2}$$ The main problem with the Purkiss Principle, most of the time, is to prove that the minimum found is global, making this part of the answer only a partial answer.
Another possibility would have been to introduce spherical coordinates, thereby eliminating the equation $a^2+b^2+c^2=3$ :
$$
a = \sqrt{3}\cos(\theta)\cos(\phi) \qquad b = \sqrt{3}\cos(\theta)\sin(\phi) \qquad c = \sqrt{3}\sin(\theta)
$$
But this destroys the beautiful symmetry in the first place. And it doesn't help much in the second place, because
the partial derivatives to $\theta$ and $\phi$ are too messy to determine the values where they are zero, which
would be required for establishing a minimum (at least MAPLE - as steered by me - refuses to do the job).
EDIT. Which is a bit too pessimistic view; one can do something with those spherical coordinates:
$$
a = \sqrt{3}\cos(\theta)\cos(\phi)=1 \quad , \quad b = \sqrt{3}\cos(\theta)\sin(\phi)=1 \quad , \quad c = \sqrt{3}\sin(\theta)=1 \\
\Longrightarrow \quad \theta = \arcsin(1/\sqrt{3})=\arctan(1/\sqrt{2}) \quad \Longrightarrow \quad \phi = \arcsin(1/\sqrt{2})=\arctan(1)=\pi/4
$$
Substitution of the spherical coordinates in our function $f$ defines $\;g(\phi,\theta)=f(a,b,c)-3/2\;$.
After some tedious calculations (or rather trusting MAPLE :-) for $(\phi_0,\theta_0) = \arctan(1,1/\sqrt{2})$ :
$$
g(\phi_0,\theta_0) = 0 \quad ; \quad
\left[ \begin{array}{c} \frac{\partial g}{\partial \phi} \\ \frac{\partial g}{\partial \theta} \end{array} \right](\phi_0,\theta_0) = 0 \quad ; \quad
\left[ \begin{array}{cc} \frac{\partial^2 g}{\partial \phi^2} & \frac{\partial^2 g}{\partial \phi \, \partial \theta} \\
\frac{\partial^2 g}{\partial \theta \, \partial \phi} & \frac{\partial^2 g}{\partial \theta^2} \end{array} \right](\phi_0,\theta_0) =
\left[ \begin{array}{cc} 3 & 0 \\ 0 & 3 \end{array} \right]
$$
Proving that there is indeed a local extreme at $\;g(\phi_0,\theta_0)\;$ and that this extreme is a minimum.
The problem is, again, to prove that the minimum found is global, making this part of the answer a partial answer as well.

- 17,070
-
2(+1) but The Purkiss Principle is a Heavy Artillery .. :) .. This proof seems to solve with only AM-GM, Cauchy Schwarz and a bit of single variable calculus – r9m Apr 15 '14 at 15:26
-
You're quite right, but I haven't found the time (yet) to double check Hawk's proof. If it is correct, then it makes it a better answer, without doubt. So maybe you will see another (+1) soon and maybe it's mine :-) – Han de Bruijn Apr 19 '14 at 11:50
By weighted AM-HM inequality, $\sum\limits_{cyc}\dfrac{\dfrac{a\cdot a^2}{a^2+b^2}}{3}\ge \dfrac{3}{\sum\limits_{cyc}\dfrac{a^2(a^2+b^2)}{a}}$
Here we consider the weights to be $a^2$,$b^2$ and $c^2$ respectively.
From the inequality above, we can have $\sum\limits_{cyc}\dfrac{a\cdot a^2}{a^2+b^2}\ge \dfrac{9}{\sum\limits_{cyc}\dfrac{a^2(a^2+b^2)}{a}}$
Now, if we can show $a(a^2+b^2)+b(b^2+c^2)+c(c^2+a^2)\le 6$ we will be done.
We can have, $a(3-c^2)+b(3-a^2)+c(3-b^2)\le 6 \implies 3(a+b+c)-(ac^2+ba^2+cb^2)$
From Cauchy-Schwarz inequality, we can have, $3(a+b+c)\le 9$
Now, we are left to prove, $ac^2+ba^2+cb^2\ge 3$
Again, using weighted AM-HM inequality,$ac^2+ba^2+cb^2\ge \dfrac{9}{\frac{ac}{c}+\frac{ab}{a}+\frac{bc}{b}}$
Now, $\dfrac{ac}{c}+\dfrac{ab}{a}+\dfrac{bc}{b}= a+b+c\le 3$
Hence, we are done.
I still believe there exists a simpler solution by a further application of Chebyshev Inequality.

- 6,540
-
1The last inequality seems to be incorrect, if we take $a=b=\frac12$ and $c=\sqrt{\frac{5}{2}}$ then $\frac{c^2}{a} = 2.\frac52 = 5 > 3$ – r9m Apr 17 '14 at 14:48
-
-
How $a^2+b^2+c^2=3$, then is equivalent to proof that $$\frac{a^3}{3-a^2}+\frac{b^3}{3-b^2}+\frac{c^3}{3-c^2}\geq\frac{3}{2}$$ Now the function $f(x)=\frac{x^3}{3-x^2}$ have $f''(x)>0$ in $(0,\sqrt{3})$ then $f(x)$ is convex in this interval, then:
$$f(a)+f(b)+f(c)\geq3f\left(\frac{a+b+c}{3}\right)$$
how $(a+b+c)^2=a^{2}+b^{2}+c^{2}+2ab+2ac+2bc$, for the Rearrangement Inequality, $2ab+2ac+2bc\leq 2a^2+2b^2+2c^2$, then $(a+b+c)^2=a^{2}+b^{2}+c^{2}+2ab+2ac+2bc\leq 3(a^2+b^2+c^2)=9\Longrightarrow a+b+c\leq3$
A hint for conclude the proof is use the "smoothing principle", ($a=b=c$), so in this case $f(a)+f(b)+f(c)$ is minimized, (but I do not understand this topic), and then:
$$f(a)+f(b)+f(c)\geq 3f(1)=\frac{3}{2}\geq3f\left(\frac{a+b+c}{3}\right)$$ God bless

- 4,212
-
Yes @Macavity thank for you comment, you is correct, I now corrected the proof. – AsdrubalBeltran Apr 14 '14 at 04:23
-
You're missing the problem - look at the denominator of the first term in LHS. It is $a^2+b^2 = 3-c^2 \neq 3-a^2$. – Macavity Apr 14 '14 at 08:17